Dark Arts 101: Be rigorous, on average

I’m read­ing Ge­orge Steiner’s 1989 book on liter­ary the­ory, Real Pres­ences. Steiner is a liter­ary the­o­rist who achieved the trifecta of hav­ing ap­point­ments at Oxford, Cam­bridge, and Har­vard. His book demon­strates an im­por­tant Dark Arts method of ar­gu­ment.

So far, Steiner’s ar­gu­ment ap­pears to be:

  1. Hu­man lan­guage is an un­de­cid­able sym­bol-sys­tem.

  2. Every sen­tence there­fore car­ries with it an in­finite amount of mean­ing, the ac­cu­mu­la­tion of all con­no­ta­tions, con­texts, and his­tor­i­cal as­so­ci­a­tions in­voked, and in­voked by those in­vo­ca­tions, etc. Alter­nately, ev­ery sen­tence con­tains no mean­ing at all, since none of those words can re­fer to things in the world.

  3. The mean­ing of a sen­tence, there­fore, is not finite or an­a­lyz­able, but tran­scen­dent.

  4. The tran­scen­dent is the search for God.

  5. There­fore, all good liter­a­ture is a search for God.

The crit­ics quoted on the back of the book, and its re­views on Ama­zon, praise Steiner’s rigor and learn­ing. It is im­pres­sive. Within a sin­gle para­graph he may show the re­la­tion­ship be­tween Homer, 12th-cen­tury the­olog­i­cal works, Racine, Shake­speare, and Schoen­berg. And his care and pre­ci­sion with words is ex­em­plary; I have the im­pres­sion, even when he speaks of mean­ing in mu­sic or other qualia-laden sub­jects, that I know ex­actly what he means.

He was in­tel­li­gent enough to trace the prob­lems he was grap­pling with out past the edges of his do­main of ex­per­tise. The key points of his ar­gu­ment lie not in liter­ary the­ory, but in in­for­ma­tion the­ory, physics, ar­tifi­cial in­tel­li­gence, com­putabil­ity the­ory, lin­guis­tics, and trans­finite math.

Un­for­tu­nately, he knows al­most noth­ing about any of those fields, and his lan­guage is pre­cise enough to be wrong, which he is when he speaks on any of those sub­jects. How did he get away with it?

An­swer: He took a two-page ar­gu­ment about things he knew lit­tle about, spread it across 200 pages, and filled the gaps with tan­gen­tial state­ments of im­pres­sive rigor and thor­ough­ness on things he was ex­pert in.

For ex­am­ple, the first chap­ter dis­cusses, with per­haps a hun­dred refer­ences, his opinion that the best art crit­i­cism is art made in re­sponse to art, his the­ory that good art is always de­rived from ear­lier art, and ob­ser­va­tions on the et­y­mol­ogy of words; but most es­pe­cially his con­ster­na­tion at the hun­dreds of thou­sands of ar­ti­cles, books, and talks on liter­a­ture pro­duced yearly by peo­ple who are not pro­fes­sors at Cam­bridge, Oxford, or Har­vard. Then on page 36, he says,

The posit­ing of an opinion about a painter, poet or com­poser is not a falsifi­able pro­ceed­ing.

I think this is the only sen­tence in the chap­ter that is a cru­cial part of his ar­gu­ment. But in­stead of en­gag­ing with the body of liter­a­ture on what falsifi­able means, whether hu­man lan­guage is falsifi­able in gen­eral (is “Ben is tall” falsifi­able?), and what falsifi­a­bil­ity has to do with the com­mu­ni­ca­tion of in­for­ma­tion, Steiner lowlights this sen­tence with its syn­tac­tic sim­plic­ity, and nearly buries it in com­plex, learned sen­tences about Dante, Mozart, and hermeneu­tics. We dive back into lit­s­peak, to emerge again into his ar­gu­ment on page 61:

All elu­ci­da­tion and crit­i­cism of liter­a­ture, mu­sic and the arts must op­er­ate within the un­de­cid­abil­ity of un­bounded sign-sys­tems… Talk can be nei­ther ver­ified nor falsified in any rigor­ous sense.

Here one should ask: If hu­man lan­guage is re­cur­sively enu­mer­able, why don’t peo­ple un­der­stand sen­tences gen­er­ated by con­text-free gram­mars for English when they go past one level of re­cur­sion (“The mouse the cat the dog chased chased squeaked”)? And doesn’t “un­de­cid­able” mean “there ex­ists at least one un­de­cid­able sen­tence” rather than “all sen­tences are un­de­cid­able”?

But one does not; one goes on to Steiner’s opinions of Tols­toy’s opinions of King Lear. It will be nearly an­other 20 pages be­fore we hit the next key point in his ar­gu­ment, which re­lies on his not know­ing that one can com­pute the sum of some in­finite se­ries. The bulk of the first 90 pages [1] is im­pres­sive dis­plays of learn­ing which fill in the vast spaces be­tween the points (al­most liter­ally) of his ar­gu­ment.

Un­less a reader pays close enough at­ten­tion to catch these brief ven­tures out­side Steiner’s ar­eas of ex­per­tise, he will come to the end of the book with (A) a sum­mary of Steiner’s ar­gu­ment, and (B) the strong im­pres­sion that the state­ments in the book were learned and rigor­ous. And thus, the ar­gu­ment car­ries.

ADDED: After look­ing at the long sec­tion on the im­pos­si­bil­ity of mean­ing that be­gins around page 90, it seems Steiner is not try­ing to ar­gue points 1 and 2 at all when he refers to them in chap­ters 1 and 2. He is merely fore­shad­ow­ing. On p. 102-103 we reach the heart of his defense of points 1 and 2, which is to say “Wittgen­stein said so.” I’m afraid that, if I finish the book, I might find a bet­ter sum­mary of the method here to be that the key points of his ar­gu­ment are defended only by ap­peals to au­thor­ity.


[1] This pat­tern breaks down around page 90, where Steiner be­gins a long spiral into his cen­tral the­sis.